Transformation Directorate

COVID-19 Clinical Risk Assessment Tool

Owner

The study was commissioned in 2020 by the Chief Medical Officer for England. The QCovid calculation engine is owned by Oxford University Innovation Limited.

The calculation engine code is licensed under the AGPL-v3 licence, which allows for free use of the software for modification, distribution, patent use, private use and commercial use.

Background

In order to manage the UK’s pandemic response, clinicians needed to understand when patients were at risk from catching COVID-19, or following a positive PCR test, hospital admission or death. Existing research around cardiovascular disease, diabetes and all cause mortality allowed researchers to build on prior work. Risk calculators designed for other conditions (QRisk3-2017, QMortality, QAdmissions) have been provided with open source licences that made it possible for them to be adapted to new applications.

Situation

Death records, hospital admissions data, and COVID-19 testing results allowed researchers to develop an appropriate model of hospitalisation and mortality. That model has also been used to create a risk stratification tool for the UK’s adult population. Individual and population risk assessment was deployed to mitigate both occupational exposure and target vaccines.

Aspiration

  • A robust and scalable clinical tool that could identify high priority COVID-19 patients most in need of early intervention to prevent hospitalisation and death.
  • The ability to nationally prioritise patients for COVID-19 vaccination

Solution and impact

The QResearch database is hosted at Oxford University. It contains anonymised data from GP and hospital records, plus COVID-19 test results and death registries. This data allowed for a tool (QCovid) to be developed specifically for the COVID-19 pandemic, taking into account a patient’s relevant risk factors including ​​age, ethnicity, gender, social deprivation, underlying medical conditions and pre-existing treatments. It provided accurate risk outcomes for patients during the first pandemic wave, validated by the Office for National Statistics (ONS) . NHS Digital then built a platform to securely include a broader set of data (including demographic information) and apply it at scale. Deployment of the tool nationally allowed for identification of people most at risk (who were clinically extremely vulnerable) so that they could be shielded from exposure and vaccinated quickly.

As the pandemic continued the QCovid model was updated to include vaccination status and background infection rate. The team also responded to specific patient groups, including treatment information for the cancer community and those with a variety of other conditions. The clinical tool was developed with extensive user testing and feedback, and relevant information was provided through the QCOVID website and NHS Digital. The deployment platform was then used to help address health inequalities. Around 1.7 million people, most of whom were extremely vulnerable (and would not be alive without risk assessment) were identified and vaccinated against COVID-19.

Functionality

The QCovid tool has been:

  • developed throughout the pandemic to provide a multivariable model that calculates a cumulative risk score for an individual
  • registered by the Medicines and Healthcare products Regulatory Agency (MHRA) as a Class 1 medical device
  • implemented as a stand-alone cloud-based service
  • utilised to provide clinicians with both absolute COVID-19 risk scores, and risk scores relative to a patient’s demographic (age, sex registered at birth, vaccination status)
  • deployed nationally for pandemic management

An ability to prioritise vulnerable patients and treatment options has informed national policy, and had an enormous impact for risk management of the pandemic and the healthcare sector as a whole.

The open source nature of the risk calculator allowed for an extremely high level of public and international scrutiny, which helped to support external validation, assurance, collaboration, and rapid roll out.

Capabilities

  • QCovid is a ‘living’ risk prediction model that can be updated by researchers at the University of Oxford to take into account emerging information about coronavirus.
  • QCovid builds on a history of other risk calculators to improve our ability to assess and manage both individual and population health risks in general
  • Its open source nature allows any part of the healthcare system to reuse the calculator as needed

Scope

The QCovid approach (both code and methodology) are open, public and reproducible. Its use throughout the pandemic represents further development of open source risk calculators at an unprecedented scale. The method implemented here can be used for rapid response to pandemics of the future, and general application to other health issues, such as diabetes.

Key learning points

  • The quantity and accuracy of information held in the healthcare system was key to creating a useful model, but limitations in the data translated directly into limitations in the model.
  • The degree to which the model and platform are open sourced are subject to ongoing discussion, with questions around the sustainability of the code leading to a more conservative approach beyond the core calculation engine.

Digital equalities

Teams at both Oxford and NHS Digital actively used the QCovid model to help address health inequalities made worse by COVID-19. The underlying data went through rigorous quality control to ensure representative coverage of multiple protected characteristics. Precautionary approaches were taken for patients with missing demographic data, especially where ethnicity data was involved. The recognition of ethnicity as a risk factor was described by the Runnymeade Trust as a ‘watershed moment’, and ‘enormously welcome’. Local Authorities with the most patients recognised as at high risk, and added to shielding efforts, were those with the high intersectionality of disadvantage (e.g. Birmingham, Newham, Tower Hamlets).

Give us feedback

Open Source Digital Playbook feedback survey

Page last updated: September 2022